Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › DYI: Virtual grounds with differential regulators
New Posts  All Forums:Forum Nav:

DYI: Virtual grounds with differential regulators

post #1 of 46
10/2/11 at 2:56pm
Thread Starter 

Since wall-warts don't have center taps, the typical DIY phone amp has a half-wave rectifier assigned to each voltage regulator. This works, but does not seem very robust with each regulator being fed one at a time.  One could use an internal power transformer but this adds cost and casework, plus probably violating electrical standards. I am interested in other solutions.

 

Consider a circuit like this:-

virt1.jpg

 

I have seen some very good write-ups about virtual grounds, especially in the Cmoy context.  But Cmoy is constrained by being portable with a single battery.  My proposed circuit is less so because the regulators compensate for uneven voltages. The low-efficiency of the split R-C pairs is a concern but there is no battery to drain.

 

A quick breadboard using 470uf, 1.3k gives about 9.5vdc, with 100mv ripple, across C1. So the amp ICs would get about +-6vdc. I am using a buffered amp with no output series resistors so I assume that I can use a lower voltage that an unbuffered one.

 

Q: Has anyone tried such a power supply variation. Do you see any pitfalls.  I plan to use 2200uf for C1,C2 but what value should I use for R1,R2?

 

Reply
post #2 of 46
10/2/11 at 7:39pm

This isn't fundamentally different from any other design of virtual ground.

 

The perturbation of the ground is dependent on the regulator gain and the impedance of the splitter R1, R2. You can improve it by reducing the resistances R1, R2 but at the cost of wasted power. You could quantify the performance either experimentally or in simulation, but you'll need good test gear or simulation skills

 

Better to use a half-wave rectifier, where the resistors become a piece of wire, and solve the other problems which are around noise created by the charging spikes which are more pronounced than in a full-wave system. Half-wave rectifier supplies aren't really any less 'robust' than full-wave ones.

 

w

Reply
post #3 of 46
10/2/11 at 8:38pm
Thread Starter 
Quote:
Originally Posted by wakibaki View Post

This isn't fundamentally different from any other design of virtual ground.

 

I agree it is fundamental but there seems to be at least 2 types .. passive and active. Intuition tells me that an active design decouples the amplifier power signals from the crap on the input side, whether it is a half-wave or a virtual circuit. In other words, if the regulators are set to +-5vdc relative to the amplifier common, then the "ground" must be held constant.

 

Of course, this is all conjecture on my part so I welcome comment.

Reply
post #4 of 46
10/2/11 at 9:01pm

The regulators are referenced to ground. If ground varies, so will the voltage from the regulators.

They will not regulate your virtual ground.

Reply
post #5 of 46
10/2/11 at 9:27pm
Thread Starter 
Quote:
Originally Posted by Avro_Arrow View Post

The regulators are referenced to ground. If ground varies, so will the voltage from the regulators.

They will not regulate your virtual ground.


So far, there have been 2 replies that are not encouraging. But I remain unconvinced so I suppose I will have to test it. Wish me luck or something.
Reply
post #6 of 46
10/3/11 at 2:00am

Just try it, you will se that it works since the mean value of audio signal is zero. But only if LM317 and LM337 (or positive and negative supply) current is the approx. the same. R1 and R2 then covers difference between positive and negative supply current.

Much better idea is to regulate voltage from wall-wart and then use rail splitter.

Reply
post #7 of 46
10/3/11 at 6:09am


Quote:

Originally Posted by fubar3 View Post


I remain unconvinced so I suppose I will have to test it. Wish me luck or something.

Yes, but how do you intend to test it? You need to run the same amplifier with a genuine dual rail supply and with your regulator virtual earth and measure the performance with a variety of test signals. You can try RMAA, but whether it will be sufficiently discriminating is the question. Crosstalk is the principal problem with virtual grounds. This is a rail splitter. What I'm trying to get across is that rail splitters are intrinsically inferior to a genuine dual-rail supply, otherwise no-one would bother with a dual-rail supply. Do you really imagine no-one else has thought of trying 2 linear regulators back-to-back with a a bridge rectifier and without a centre-tapped transformer? In order to answer your question I had to work out why a half-wave rectified supply is superior (although not in terms of radiated or induced noise).

 

Good luck.
 

w

Reply
post #8 of 46
10/3/11 at 8:26am

I took the dogs for a walk, and as is frequently the case, while I was out walking a penny dropped.

 

The big trouble with this arrangement is that if the difference in return currents is sufficient, then the junction of R1, R2 can move sufficiently close to one of the rails that one of the regulators drops out of regulation, which I suspect may result in it latching in that fault condition.

 

It may even be the case that in some instances it is difficult to get the circuit to fire up correctly

 

w

Reply
post #9 of 46
10/3/11 at 10:13am

 

Quote:
Originally Posted by wakibaki View Post

Crosstalk is the principal problem with virtual grounds....


This is interesting, can you explain why my portable amp ("CMOY") has crosstalk over 85dB (through 10Hz to 20kHz). Power supply is just Li-ION cell, two 470uF caps and two 1kOhm resistors.

So as you can see, virtual ground DOES NOT cause bad crosstalk. Only badly designed virtual ground can cause that.

 

Latch-up can be problem, but lowering of R1 and R2 isn't problem. (1k in Li-ION powered CMOY succesfully prevents larching)


Edited by AmarokCZ - 10/3/11 at 10:16am
Reply
post #10 of 46
10/3/11 at 3:51pm


 

Quote:
Originally Posted by AmarokCZ View Post

 

This is interesting, can you explain why my portable amp ("CMOY") has crosstalk over 85dB (through 10Hz to 20kHz). Power supply is just Li-ION cell, two 470uF caps and two 1kOhm resistors.

So as you can see, virtual ground DOES NOT cause bad crosstalk. Only badly designed virtual ground can cause that.

 

Latch-up can be problem, but lowering of R1 and R2 isn't problem. (1k in Li-ION powered CMOY succesfully prevents larching)

Measured how? Did you take this number from an advert? Most 3.5 stereo jacks have a crosstalk figure worse than this, -70dB or worse. -85dB is the kind of figure you might obtain with the output unloaded.

 

I think you are mistaken. A shared virtual ground will always increase the crosstalk compared to a conventional star ground, because the higher output impedance of the virtual ground is shared between the channels. This shared impedance means that the channels interact in a way which would be impossible otherwise. It is beyond argument that rail-splitters add this shared increase in impedance.

 

How could it be any different?
 

I am not interested in arguing with you, stirring up your prejudices, or insulting your equipment. I am merely explaining the situation from an engineering perspective to fubar3 for his information. If you have an alternate and coherent engineering-based understanding of the situation, please present it.

 

Unnecessary use of bold and SHOUTING do nothing for your case.

 

Perhaps you can explain why power amplifiers (other than amateur headphone amplifier designs) almost without exception use a dual-rail supply instead of a rail splitter? When single rails are employed from necessity, the amplifier will be AC (capacitor) coupled.

 

w

 


Edited by wakibaki - 10/3/11 at 3:53pm
Reply
post #11 of 46
10/3/11 at 9:58pm
Thread Starter 
Quote:
Originally Posted by wakibaki View Post

Quote:

Yes, but how do you intend to test it?

I would apply a DC voltage to the amplifier input.  Measure voltages at various points while the amplifier output is offset.

 

>>Do you really imagine no-one else has thought of trying 2 linear regulators back-to-back with a a bridge rectifier and without a centre-tapped transformer? In order to answer your question<<

I would be surprised if no-one else considered the idea.

 

>>I had to work out why a half-wave rectified supply is superior (although not in terms of radiated or induced noise).<<

Much appreciated.
 

Reply
post #12 of 46
10/3/11 at 10:32pm
Thread Starter 
Quote:
Originally Posted by wakibaki View Post

I took the dogs for a walk, and as is frequently the case, while I was out walking a penny dropped.

 

The big trouble with this arrangement is that if the difference in return currents is sufficient, then the junction of R1, R2 can move sufficiently close to one of the rails that one of the regulators drops out of regulation, which I suspect may result in it latching in that fault condition.

 

It may even be the case that in some instances it is difficult to get the circuit to fire up correctly

 

w

Yes, I understand how that can happen.  But before this rail-split experiment, I already prototyped a "surfboard amp" with 49610 buffers for gain-stage isolation and LO-Z phones.  But I added limiters in the supply for phone protection. It worked great on my ears but was prone to power-up instability because the LM337 is much slower than the LM117 or because of the amp design. The latch up does not happen if the volume is zero or the phones are unplugged.  So now I am interested in variant supplies.
 

Oh, when the latch up occurs, one of the LM337 takes the heat. But I need to build a simpler supply that somehow limits the max juice through the phones.

 

Here is one for LM117.. but it needs a transformer.

limit1.gif

Reply
post #13 of 46
10/3/11 at 10:44pm

just use a dual rail supply, this has been done to death and proved inadequate 

Reply
post #14 of 46
10/4/11 at 1:43am

wakibaki: it was measured with 16R headphones as load and measured with AC voltmeter/uV meter.

Your statement that most 3.5mm Jacks have "-70dB or worse", this statement is ... well ... no offense, but it's just total nonsense.

 

And yes I can explain why most (well ... not most, but just every) power amps use dual-rail. It's because there is no reason to use anything else than dual-rail.

Dual rail is simpler (just transforem with more windings vs. electronic splitter), more reliable, doesn't cause any unwanted interactions between channels and to be honest I can't really imagine (for example) 1KW supply that would use rail splitter.

Even at DIY there is no reason to use splitter, but if fubar3 has no other choice (consedering his concerns), this is the only way to power that amp.

 

I can agree with you that rail-splitters cause some problem, but when used in low power/current amps (e.g. portable headphone amp) it's not that horrible.

 

BTW that shared impedance in "CMOY passive PSU" is product of capacitors ESR, so it's quite simple to reduce this impedance.

 

EDIT: "Jack crosstalk" problem: few days ago I finished headamp a I used very cheap 0.35€ 3.5mm jack (see this photo).

When measuring parameters I connected jack splitter (1x male to 2x female adaptor) to that cheap 3.5mm jack. In one female plug there was AKG K240 connected as load and second female plug was connected to voltmeter.

And resulting crosstalk was: -101dB@100Hz, -105dB@1kHz and -96dB@15kHz. Crosstalk directly on amplifier PCB was approx. the same.

So please explain me where did you get the idea that "Most 3.5 stereo jacks have a crosstalk figure worse than this, -70dB or worse."

 

 

 

fubar3: if you really want splitter based PSU use circuit similar to the one at your first post (after testing latch-up immunity). But I would reccomend dual-rail supply too.


Edited by AmarokCZ - 10/4/11 at 2:34am
Reply
post #15 of 46
10/4/11 at 7:18pm
Thread Starter 
Quote:
Originally Posted by qusp View Post

just use a dual rail supply, this has been done to death and proved inadequate 


Been there, done that.. I built the typical half-wave rectifiers feeding regulators. Prelimary results were good but I was afraid of smoking the phones with my experiments so I added current-limiting regulators. But the negative voltage came up only after the positive. This caused the amp to sometimes latch on power-up or prevented the limiter from charging. I did more hacking with slow-turn on circuits from the LM317 datasheet examples.  I can personally use that contraption, but it is too scary.. even for me.

S2.jpg

 

Reply
New Posts  All Forums:Forum Nav:
  Return Home
Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › DYI: Virtual grounds with differential regulators